An AI-based face recognition security method

Abstract

The present application relates to the technical field of security monitoring, in particular to a face recognition security method based on AI, comprising the following steps: recognizing the channel based on the monitoring entrance and exit, collecting and analyzing multispectral face images, screening feature points and skin color areas, optimizing spatial feature matching and structure distribution, fusing head movement and facial behavior parameters, judging structure consistency and linkage features, and outputting face recognition mapping results. In the present application, face detail level discrimination is realized through multispectral information fusion, based on regional pixel distribution and structure sequence, combined with dynamic spatial trajectory analysis and abnormal motion correction, continuous spatial features and time sequence behavior parameters are matched in parallel, behavior linkage and facial structure features are screened synchronously, and the multi-parameter joint recognition strategy in the dynamic scene reduces the interference caused by environmental changes and human intervention, improves the reliability of target identity determination in complex places, and strengthens the practicality and real-time recognition capability of multi-source scene.

Description

TECHNICAL FIELD

[0001] The application relates to the security monitoring technical field, in particular to a face recognition security method based on AI. BACKGROUND

[0002] The security monitoring field relates to an information acquisition and processing system for real-time sensing, recognizing and recording specific areas, objects or personnel, mainly including video monitoring, behavior analysis, image recognition, identity verification and abnormal event detection, aiming to improve the security protection capability of a place, and is widely applied to public security, traffic management, community governance, enterprise protection and the like. The traditional face recognition security method refers to acquiring a target face image by using an image acquisition device, comparing the face key feature points extracted by a feature extraction algorithm with feature templates in a face database to realize identity recognition, and verifying and recording personnel entering and leaving a place.

[0003] The prior art is based on static feature point comparison under a single spectrum, lacks the cooperative determination capability of time sequence and space parameters in a dynamic scene, is slow in responding to behavior changes and multi-scene switching, is difficult to effectively respond to continuous changes such as motion state, light disturbance and facial micro-movement, cannot realize adaptive fusion of multi-dimensional features, is prone to cause problems such as abnormal target recognition omission, disguise misjudgment and multi-source signal fusion failure, and affects the adaptation and response of security monitoring to variable places. SUMMARY

[0004] The application aims to solve the problems in the prior art and provides a face recognition security method based on AI.

[0005] In order to achieve the above-mentioned purpose, the application adopts the following technical scheme: a face recognition security method based on AI, comprising the following steps:

[0006] S1: based on a monitoring entrance and exit recognition channel, analyzing a face image pixel distribution, judging the contour definition captured by an infrared visual array, comparing the collection details of a multi-spectrum sensor, screening a feature point and a complete frame of symmetry parameters, determining a skin color region distribution, and obtaining a multi-spectrum face feature;

[0007] S2: based on the multi-spectrum face feature, calling an image recognition buffer unit, optimizing a face feature aggregation channel multi-frame image, judging the uniformity of a feature point space distribution, comparing a contour structure overlap, screening a continuous face frame with a space difference in an interval, and obtaining a space feature screening sequence;

[0008] S3: filtering sequences based on the spatial features, analyzing the three-dimensional space variation of feature points, comparing the X, Y axis position trend of continuous frames, judging the Z axis trajectory coherence, filtering the position change stable feature points, performing correction on the trajectory mutation points, and obtaining a three-dimensional trajectory sequence;

[0009] S4: based on the three-dimensional trajectory sequence, calculating the head rotation angle change, judging the face micro-motion speed recorded by the local motion capture unit, comparing the action parameters and the feature point space coordinate change mode, filtering the data with the optimal matching degree, and obtaining the action linkage matching data.

[0010] The application improves that the multispectral face features include spectral distribution parameters, texture feature quantities and skin color block information, the spatial feature filtering sequence includes spatial consistency indicators, structure aggregation labels and interframe stable segments, the three-dimensional trajectory sequence includes coordinate trajectory data, coherence labels and corrected trajectory sets, and the action linkage matching data includes behavior linkage parameters, action recognition numbers and linkage matching relationship groups.

[0011] The application improves that the multispectral face feature acquisition step specifically comprises:

[0012] S111: based on the monitoring entrance and exit identification channel, analyzing the pixel distribution of the face image data, for each block of the image sequence, calculating the distribution density of the edge continuity of each pixel block by counting the change level of the gray gradient, judging the edge definition change trend of each region in the infrared visual array capture region, and obtaining the infrared contour definition distribution amount;

[0013] S112: based on the infrared contour definition distribution amount, comparing the brightness histogram distribution in the images obtained by the multispectral synchronous sensor array at different time points, counting the texture direction density of each region, dividing the face key region by analyzing the aggregation and closed characteristics of the boundary pixels, and obtaining the face structure aggregation data;

[0014] S113: based on the face structure aggregation data, filtering the feature point coordinates and symmetry mapping set in the image frame, performing multi-channel fusion superposition, and obtaining the multispectral face features.

[0015] The application improves that the spatial feature filtering sequence acquisition step specifically comprises:

[0016] S211: based on the multispectral face features, analyzing the arrangement of each feature point on the horizontal and vertical coordinate axes, judging whether the spatial distribution has the balance and symmetry characteristics of the face structure, filtering the image frames with consistent spatial distribution rules, and obtaining a set of symmetrically distributed image frames.

[0017] S212: Based on the symmetric distribution image frame set, the spatial trend of the same region boundary line and the connection relationship are compared, the continuity of the key contour line between frames and the boundary connection situation are calculated, the frame images with consistent contour connection mode are screened, and a contour continuous image frame set is obtained.

[0018] S213: Based on the contour continuous image frame set, the coordinate change of each frame feature point is judged, the moving trend of the key point between adjacent image frames is analyzed, the image frame sequence with stable feature point motion rule is screened, and a spatial feature screening sequence is obtained.

[0019] The application improves that the acquisition step of the three-dimensional trajectory sequence is specifically:

[0020] S311: Based on the spatial feature screening sequence, the plane coordinates of each feature point in the X axis and the Y axis in the continuous frame are extracted, the coordinate change amplitude of the same feature point in adjacent frames is compared, the change trend thereof in the time sequence is identified, and an XY axis trend sequence is established.

[0021] S312: Based on the XY axis trend sequence, the distribution of the Z axis coordinates of each group of feature points in the continuous frame is detected, the continuous coordinates of the Z axis in the time sequence are combined, the mutation segment is identified, the interpolation smoothing processing is performed, the trajectory change parameter of each group of feature points in the three-axis space is acquired, and the abnormal trajectory is corrected, and a Z axis trajectory correction set is obtained.

[0022] S313: Based on the Z axis trajectory correction set, in combination with the trajectory continuity of the feature points in the three-axis space, the trajectory segment with balanced space change amplitude is screened, the continuous space coordinates thereof in the time sequence are acquired, and a three-dimensional trajectory sequence is obtained.

[0023] The application improves that the acquisition step of the action linkage matching data is specifically:

[0024] S411: Based on the three-dimensional trajectory sequence, the X axis and Y axis position data of each frame feature point are selected, the continuous time sequence of the Z axis trajectory is combined, the head rotation angle change amount is data extracted, the angle difference of each frame face region displacement trajectory is calculated, the posture transformation angle under each time sequence is matched with the corresponding feature point coordinate offset amount to establish an item-by-item matching relationship, and angle displacement matching data is obtained.

[0025] S412: Based on the angle displacement matching data, the speed information of each region micro-motion is detected, the coordinate change amplitude and time interval of each region are compared, and the speed joint offset trend is calculated.

[0026] S413: Based on the rate joint offset trend, data comparison is performed on each group of micro-motion data and feature point space motion change, the correlation degree of the structure similarity parameter and the action linkage index in the combination is analyzed, data collection is performed on the combination meeting the linkage condition, and action linkage matching data is obtained.

[0027] The application improves that the step further comprises:

[0028] S5: Based on the action linkage matching data, the corresponding relationship between the numbered feature points and the behavior parameters is analyzed, the data structure consistency of each group of the recognition channel is judged, the discrimination process of the joint recognition integrated port is optimized, the group meeting the structure and behavior standards is recognized, the recognition number is marked, and a face recognition mapping result is obtained.

[0029] The face recognition mapping result comprises an identity mapping identifier, a recognition identifier, and a feature linkage mapping index.

[0030] The application improves that the face recognition mapping result is obtained by the following steps:

[0031] S511: Based on the action linkage matching data, the corresponding relationship between the numbered feature points and the behavior parameters is analyzed, the continuous distribution of each feature point in the three-dimensional space and the synchronism of the behavior change are judged, the corresponding group with high consistency of the structure feature and the behavior feature is screened, and corresponding distribution data is obtained.

[0032] S512: Based on the corresponding distribution data, the spatial structure relationship and the behavior change mode of the difference data group are compared, the arrangement order of the feature point combination and the corresponding rule between the behavior parameters are optimized, the combination with consistent structure and behavior is screened, and structure mapping combination data is obtained.

[0033] S513: Based on the structure mapping combination data, the linkage distribution between the feature point group and the behavior parameter group is judged, the participation proportion of each feature point and the linkage characteristics of the behavior parameter are calculated, the synchronization relationship and the mapping index are marked, and a face recognition mapping result is obtained.

[0034] Compared with the prior art, the application has the advantages and positive effects that:

[0035] In the application, the face detail level discrimination is realized through multispectral information fusion, the continuous space feature and the time sequence behavior parameter are matched in parallel based on the regional pixel distribution and the structure sequence, the dynamic space trajectory analysis and the abnormal motion correction are combined, the behavior linkage and the face structure feature are synchronously screened, the multi-parameter joint recognition strategy in the dynamic scene reduces the interference caused by the environmental change and the human intervention, improves the reliability of the target identity judgment in the complex place, and strengthens the practicability and the real-time recognition capability of the multi-source scene. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 Flow chart of main steps of the present application;

[0037] Figure 2 Flow chart of acquisition of multi-spectral facial features in the present application;

[0038] Figure 3 Flow chart of acquisition of spatial feature screening sequence in the present application;

[0039] Figure 4 Flow chart of acquisition of three-dimensional trajectory sequence in the present application;

[0040] Figure 5 Flow chart of acquisition of action linkage matching data in the present application;

[0041] Figure 6 Flow chart of acquisition of face recognition mapping result in the present application. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0043] In the description of the present application, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0044] EMBODIMENT

[0045] Please refer to Figure 1 The present application provides a technical scheme: an AI-based face recognition security method, comprising the following steps:

[0046] S1: Based on the monitoring entrance and exit identification channel, analyze the pixel distribution of the face image data, judge the face contour sharpness of the infrared visual array capture area, through comparing the image details collected by the multi-spectral synchronous sensor array, structure the segmentation of the face region under the collection time sequence, screen the frame with complete feature point coordinates and symmetry parameters, determine the skin color region distribution range, and obtain the multi-spectral facial features;

[0047] S2: Based on the multispectral facial features, calling the image recognition buffer unit, performing optimization on the multi-frame image output by the face feature aggregation channel, judging whether the spatial distribution of each frame of feature points is uniform, comparing the structural overlap between the contour boundaries of each frame, screening the face image frames with continuous structure and spatial difference within the preset interval, and obtaining a spatial feature screening sequence;

[0048] S3: Based on the spatial feature screening sequence, analyzing the three-dimensional spatial changes of the feature points between frames, comparing the change trend of the X-axis and Y-axis positions in the continuous frames, judging the continuity of the Z-axis trajectory under the time sequence, screening the feature points with stable position changes, correcting the points with sudden changes in the spatial trajectory, and obtaining a three-dimensional trajectory sequence;

[0049] S4: Based on the three-dimensional trajectory sequence, accessing the behavior parameter comparison interface, calculating the change of the head rotation angle, judging the face micro-motion speed recorded by the local motion capture unit, comparing each group of motion parameters with the change mode of the spatial coordinates of the feature points, screening the data that meet the matching requirements in structure similarity and motion linkage, and obtaining motion linkage matching data;

[0050] S5: Based on the motion linkage matching data, analyzing the corresponding relationship between the numbered feature points and the behavior parameters, judging the structural consistency of each group of data in the identity recognition channel, optimizing the discrimination process of the joint recognition integrated port, identifying the feature groups that meet the standard requirements in structure and behavior, comparing the linkage feature distribution proportion of each group, and marking each identification number, and obtaining a face recognition mapping result.

[0051] The multispectral facial features include spectral distribution parameters, texture feature quantities, and skin color block information. The spatial feature screening sequence includes spatial consistency indicators, structure aggregation labels, and inter-frame stable segments. The three-dimensional trajectory sequence includes coordinate trajectory data, continuity labels, and corrected trajectory sets. The motion linkage matching data includes behavior linkage parameters, motion recognition numbers, and linkage matching relationship groups. The face recognition mapping result includes identity mapping labels, recognition labels, and feature linkage mapping indexes.

[0052] In S1, the identification channel refers to a physical channel or a virtual detection area with face collection and identification functions for security scenarios, commonly used in access control, entrances, checkpoints, and other monitoring deployment sites. The infrared vision array refers to an array device composed of multiple infrared sensing units, used to capture infrared images of faces in low-light, backlight, or night environments, enhancing the adaptability of face detection. The multispectral synchronous sensor array refers to an integrated sensing module that can simultaneously collect visible light, near-infrared, short-wave infrared, and other multispectral images, ensuring the acquisition of multi-source face image information under different lighting and camouflage conditions. Structured segmentation refers to accurately separating the face region from panoramic or background images and further dividing it into specific functional areas (such as eyes, nose, mouth, and face shape) for subsequent precise feature analysis. Complete frames refer to image frames where key facial features (such as facial features and contours) are not obscured or missing, and the image quality meets the analysis requirements, suitable for subsequent feature extraction and comparison. Skin color region distribution range refers to the determination of skin tone and skin color distribution area boundaries within the face region through pixel color analysis, which helps to assist in investigating obstructions or disguises and improves recognition robustness.

[0053] In S2, the image recognition buffer unit refers to a buffer area or module for temporarily storing and managing multiple frames of face image data, facilitating continuous processing and screening, and improving system real-time performance and stability. The face feature aggregation channel refers to a dedicated processing channel that collects, integrates, and aligns multiple frames of collected face features (such as key points and region segmentation results). The uniformity of distribution refers to whether the distribution of face key feature points (such as eyes, nose, and mouth) in the image space is regular and free of abnormal deviations, avoiding false positives caused by abnormal detection. The structure overlap condition refers to comparing the overlap degree of face key part boundaries in different frames of images to measure the consistency of multiple frames of images and exclude recognition errors caused by factors such as shaking and blurring. The spatial difference within the preset interval refers to judging whether the spatial position change of the same feature point in multiple frames of images is within the normal range defined by the system, ensuring that the selected frames have high similarity.

[0054] In S3, the three-dimensional space change refers to the position change process of the facial feature points in the X, Y, and Z three space dimensions (plane position + depth), reflecting the facial action or posture change; the change trend refers to analyzing the space coordinate change track of the feature points in the continuous multiple frames, judging whether it presents a certain trend such as stability, linearity, periodicity, or abnormal deviation; the continuity refers to measuring the continuity and stability of the three-dimensional track in the frame sequence, avoiding abnormal phenomena such as mutation, jump point, etc., and ensuring the accuracy of feature tracking; the feature point with stability refers to the facial feature point with small space position change amplitude, smooth track, and not easy to be disturbed in multiple frames of data, which is suitable for being used as the basis for identity discrimination; the space track refers to the motion path formed by the feature points in the three-dimensional coordinate system with time passing, which is used for dynamic analysis and identity recognition; the track correction refers to adjusting the feature point track with abnormal fluctuation or mutation by using algorithm smoothing, interpolation, or abnormal point elimination, etc., to ensure the consistency of the overall track.

[0055] In S4, the behavior parameter comparison interface refers to the data interface or processing unit in the system for accessing and processing the parameters related to facial actions (such as head rotation, expression change, etc.); the local motion capture unit is a collection module specially used for high-frequency collection and recording of facial micro-motions (such as mouth corner twitching, eyebrow lifting, etc.), which improves the dynamic behavior analysis capability; the facial micro-motion speed refers to measuring the speed characteristics of facial local (such as eyes, mouth, eyebrows, etc.) motion through time series analysis, which is an important basis for distinguishing living beings and disguises; the change mode refers to describing the change law between action parameters and feature point space coordinates with time passing, such as synchronous, delayed, and sudden change modes; the structural similarity refers to quantifying the similarity of key structural features between data samples, which is commonly used for screening action data of the same identity or high correlation; the motion linkage refers to analyzing whether there is a cooperative change relationship between head and facial local actions, which is used to determine the naturalness of the action or whether there is a fake behavior; the matching requirement refers to the minimum standard of the correlation degree between structural features and behavior features in the identification process, and data below the standard will be excluded.

[0056] In S5, the numbered feature points refer to assigning unique numbers to the key points collected for each face image, which facilitates accurate tracking and identification of corresponding points when multiple frames and multiple parameters are fused; the corresponding relationship refers to the one-to-one corresponding and mutually associated relationship between the feature point data and the behavior parameter data, which facilitates joint analysis and comprehensive determination of multiple parameters; the identity recognition channel refers to a data processing channel used for making a final identity recognition decision for candidate face data, which is a core discrimination link; the structural consistency refers to the consistency of the same candidate identity in each feature dimension under multiple parameter fusion, and high structural consistency indicates high identity determination reliability; the joint recognition integrated port refers to an interface or data endpoint that integrates multiple-dimensional recognition results, behavior actions, spatial features and other multi-source data for comprehensive output and determination; the feature group meeting the standard requirements refers to a feature point combination that passes all structural and behavior screening standards, and has high reliability and unique identity feature group; and the labeled recognition number of each group refers to assigning a unique recognition number to each identity candidate group for subsequent output and management.

[0057] Referring to Figure 2 , the acquisition steps of the multi-spectral face features are specifically as follows:

[0058] S111: Based on the monitoring entrance and exit recognition channel, the pixel distribution of the face image data is analyzed, for each block of the image sequence, the distribution density of the edge continuity of each pixel block is calculated by counting the change level of the gray gradient, the edge definition change trend of each region in the infrared visual array capture region is judged, and the infrared contour definition distribution quantity is obtained;

[0059] The pixel-level analysis is performed on the face image sequence collected by the infrared visual array. Each frame of image is divided into a plurality of pixel block regions according to a fixed grid. Each pixel block can be set to 32 by 32 pixels in size. The change information of the gray value in each pixel block is extracted block by block. The difference value sequence between each group of adjacent pixels is counted. Whether the current pixel block has obvious edge features is judged by comparing the change amplitude of the difference value. The number of pixel pairs in each pixel block whose gray value change amplitude reaches or exceeds the set reference value is further counted. The proportion of the number in the entire pixel block is calculated. If the proportion exceeds 30%, it is determined that the region has high edge continuity. The pixel blocks satisfying the condition are indexed, and the corresponding spatial position in the entire image is recorded. The number of pixel blocks satisfying the edge continuity condition in the unit region is counted by traversing the image space to form the spatial distribution data of the edge sharpness. Whether the region is an edge sharpness region is judged according to whether the number of clear pixel blocks in each region reaches 600. All edge sharpness regions in the image are extracted and their outlines are merged. The arrangement and distribution density change of the edge sharpness regions in the image space are analyzed. The spatial boundary position of each region is recorded to form the trend map of the edge distribution. The continuous frames of the image sequence are processed. The number change of the edge sharpness regions in the same spatial position in the frames collected at different time points is compared. Whether the change rate is stable is counted. If the change proportion of the sharpness of a region in all frames is less than 5% and the area proportion of the region in the image is more than 70%, the region is determined to be a stable clear region under infrared capture. The distribution trend data of the edge outline clear region of the entire image under the infrared visual array is obtained as the infrared outline clear distribution quantity.

[0060] S112: Based on the infrared outline clear distribution quantity, the brightness histogram distribution in the images obtained by the multispectral synchronous sensor array at different time points is compared. The texture direction density of each region is counted. The face key region is divided by analyzing the aggregation and closed characteristics of the boundary pixels to obtain the face structure aggregation data.

[0061] The image brightness comparison analysis is performed on the face images collected by the multispectral synchronous sensor array at different times. The images taken at two time points are selected and grayscale processed. The brightness information in the image is extracted as a distribution matrix. The image brightness distribution is then normalized to compare the illumination changes between the images. The image brightness distribution at the two time points is then plotted as a histogram. The difference in the frequency of each brightness level in the two frames of images is calculated. The difference of all brightness levels is averaged as a comparison indicator of the overall brightness change. If the average difference exceeds 15, it is considered that there is a significant brightness difference between the images. At the same time, the area corresponding to the brightness difference is marked in the image space. After that, the image is edge-corrected. Extraction processing, statistics of texture direction information in each area, by traversing the grayscale change direction of each pixel in its adjacent area, determine whether the dominant texture direction in the area is concentrated, if the proportion of the dominant direction exceeds 80%, the area is judged as a texture direction dense area, then judge the aggregation state of the boundary pixels, and judge its closed characteristics by detecting whether the edge pixels form a closed contour. If the proportion of closed boundary pixels is greater than 60%, the area is marked as a boundary closed area. Finally, the areas in the image space that meet the requirements of significant brightness changes, concentrated texture directions, and closed boundary structures are collectively processed, and their spatial position and structural feature data are output to obtain facial structure aggregation data for subsequent structural analysis.

[0062] S113: Based on the facial structure aggregation data, the feature point coordinates and symmetry mapping set in the image frame are screened using the formula:

[0063] ;

[0064] Perform multi-channel fusion and superposition to obtain multispectral facial features ,in, Indicates the The symmetry combination factor of the feature points, Indicates the The brightness frequency band response of the feature points in the left area after structure division, Indicates the The brightness frequency band response of the feature points in the right area after structure division, Indicates the The degree of texture interference within the region, represents the spectrum collaborative fusion enhancement term, Represents the superposition of visible light spectrum reflectance measurement, represents the absorption response fusion amount in the near-infrared band, Indicates skin color area coverage mapping adjustment item, Indicates the pixel coverage composite area of ​​the skin color block area, Indicates the number of feature points.

[0065] The multi-spectrum face feature is a comprehensive expression of the face image feature after fusion operation, which is used for identity feature mapping, feature point screening, structure consistency determination and other processing procedures in the subsequent face recognition system, and is a multi-dimensional basic data index for distinguishing individual identity, detecting disguises or obstructions.

[0066] The feature points of each key region marked in the image frame sequence are extracted and divided into a number set. For each number corresponding feature point, the brightness response of the left and right symmetric regions of the face in the visible light spectrum channel is extracted 、 , the texture interference degree of the region where the point is located is measured , and the symmetry combination factor is given . The multi-channel fusion superposition is performed on each parameter according to the following formula, and the three feature points with number have the following original values:

[0067] 、 、 、 ;

[0068] 、 、 、 ;

[0069] 、 、 、 ;

[0070] The normalized value of the visible light reflection measurement superposition amount is , the normalized value of the near-infrared absorption response fusion amount is , the fusion enhancement item is , the skin color region mapping adjustment item is , and the normalized value of the skin color block pixel coverage area is .

[0071] The calculation is as follows:

[0072] The first term:

[0073] ;

[0074] The second term:

[0075] ;

[0076] Item 3:

[0077] ;

[0078] The weighted structure term is obtained by summing the above three terms:

[0079] ;

[0080] The spectral enhancement term is calculated again:

[0081] ;

[0082] The denominator is:

[0083] ;

[0084] The denominator is:

[0085] ;

[0086] The multispectral facial feature is:

[0087] ;

[0088] The results show that the multispectral facial feature image quantity GS≈6.11 calculated at present is higher than the upper limit value 4.5 of the standard reference interval, which is obtained by statistical fusion feature expression results of a large number of facial image samples, and usually distributed in the interval of 2.0~4.5, wherein GS<2.0 is marked as insufficient feature expression, and GS>4.5 represents high fusion feature integrity, stable structure matching, and active spectral response. The numerical result shows that the feature points extracted in the current image frame show high coordination in structure symmetry, regional texture, and response in multispectral channels, so the overall fusion value is significantly amplified, representing that it has the basic qualification of direct inclusion into the next stage of spatial screening and recognition comparison process, and the subsequent spatial trajectory stability analysis and dynamic behavior comparison process can be participated according to the feature quantity.

[0089] Please refer to Figure 3 , the acquisition steps of the spatial feature screening sequence are as follows:

[0090] S211: Based on the multispectral facial feature, analyze the arrangement of each feature point on the horizontal and vertical coordinate axes, judge whether the spatial distribution has the balance and symmetry characteristics of the face structure, and screen the image frames with consistent spatial distribution rules to obtain a set of symmetrically distributed image frames;

[0091] Based on the multispectral facial features, the facial feature data includes the transverse and longitudinal coordinate positions of the feature points extracted in each frame of image, the skin color region distribution, the texture line position, etc. At least 50 feature points are extracted from the face image, and the X-axis and Y-axis coordinate values of each feature point in the two-dimensional coordinate system of the image are recorded. First, a coordinate table is established independently for each frame of image. The coordinate range of all feature points in the horizontal axis direction is calculated, and the central axis of the face (such as the vertical line passing through the center point between the two eyes) is taken as the reference line to determine whether the number of feature points on the left side and the right side is equal. It is determined whether the horizontal axis coordinates of the corresponding points on the two sides are distributed in pairs about the central axis, that is, whether the symmetry difference is less than a set symmetry error threshold. For example, the threshold is set to 3 pixel points. The symmetry difference is the difference between the two corresponding points from the center line. If the proportion of all pairs of feature points that satisfy the symmetry difference less than the threshold reaches more than 80%, it is determined that the image frame has symmetry characteristics. Further, the distribution of feature points in the vertical axis direction is evaluated to record whether each point is evenly distributed or uniformly arranged in the vertical direction, and whether there is a concentrated or abnormally aggregated area. If the transverse and longitudinal coordinate distributions simultaneously satisfy the above requirements of structural balance and symmetry, the frame of image is added to the preliminary selection set. Each frame of image in the image sequence is traversed, and the above operation is repeated to select the image frame samples that satisfy the spatial symmetry structure characteristics. The symmetrically distributed image frame set is recorded by establishing an index.

[0092] S212: Based on the symmetrically distributed image frame set, the spatial trend and connection relationship of the same region boundary line are compared, the continuity of the key contour line between each frame and the boundary connection situation are calculated, the frame images with consistent contour connection mode are selected, and the contour continuous image frame set is obtained.

[0093] The spatial structure of the key facial boundary line in each frame image is evaluated in sequence. The key boundary line refers to a nose bridge contour line, a facial side edge line, a lower jaw contour line, etc. extracted from the image. First, the adjacent connection position of each pixel on the boundary line is extracted according to the pixel coordinates, and a boundary pixel continuous direction sequence is generated. Then, the extension direction difference of the boundary line in the same facial region in different frame images is compared. The direction consistency of the boundary direction is determined using the coordinate difference value. For example, the direction vector difference of the nose bridge contour line in frame 1 and frame 2 is taken. If the included angle of the two vectors is less than 10 degrees, it is determined that the directions are consistent. If the included angle is greater than 15 degrees, it is determined that the difference is obvious. Then, the boundary line connection condition is calculated. Whether the contour line is broken is determined by the connection point density. The ratio of the number of connection points to the length of the line segment is taken as the continuity reference index. If the ratio is higher than a set threshold value (for example, more than 8 continuous connection points exist per 10 pixel length), it is considered that the boundary connection is good. Further, the start and end point coordinates of the boundary line in each frame are compared. If the start and end point offset distance is less than 5 pixels between two frames, it is determined that the boundary alignment condition is good. Through the consistency of the direction vector, the connection density and the start and end coordinates of the same boundary line in all image frames, image frames that meet the requirements of the above three judgment conditions are selected, and are recorded as contour connection mode consistent frames. The contour continuous image frame set is formed by collecting.

[0094] S213: Based on the contour continuous image frame set, the coordinate change of each frame feature point is determined, the movement trend of the key point between adjacent image frames is analyzed, the image frame sequence in which the feature point motion law remains stable is selected, and the spatial feature selection sequence is obtained.

[0095] The position change of the feature point between each image frame is analyzed. The coordinates of all key feature points in each frame image are extracted in sequence, and the point position correspondence relationship with the previous frame is established. Whether there is a corresponding point number disorder or omission is determined. First, the coordinate change value of each feature point in the X-axis direction in the continuous frame is calculated. Then, the same processing is performed on the Y-axis direction. The displacement vector sequence of each feature point is obtained. Whether the average displacement distance of all feature points is within a stable change range is recorded by statistical analysis. The threshold value of the range is set to be that the maximum movement of each frame is not more than 5 pixels. If the displacement distance of a certain feature point is less than the value for three consecutive frames, it is determined that the motion trend is stable. Then, the number of points that meet the stability condition among all feature points is counted. If the proportion of the number of points to the total number of feature points is more than 85%, the frame image is recorded as a feature point motion law stable frame. The above judgment is repeated by traversing all frames. The image frames with high feature point position stability are arranged and combined to form the spatial feature selection sequence by establishing a stable frame sequence index set.

[0096] Please refer to Figure 4 The steps of obtaining the three-dimensional trajectory sequence are specifically as follows:

[0097] S311: Based on the spatial feature screening sequence, the planar coordinates of each feature point in the X axis and Y axis in the continuous frame are extracted, the coordinate change amplitude of the same feature point in the adjacent frame is compared, the change trend of the same feature point in the time sequence is recognized, and the XY axis trend sequence is established.

[0098] Each image in the sequence has been screened by the feature structure, and has the continuous and stable face feature point distribution characteristics. The X axis and Y axis values of the key feature points in the planar coordinate system are extracted for all the frame images in sequence. The coordinates of each feature point in the image are represented by two-dimensional pixel positions, for example, a certain point is (120, 250) in frame 1 and (122, 252) in frame 2. The one-to-one correspondence relationship between the frame feature point numbers is established in time sequence. The X axis and Y axis coordinates of the same number point in each pair of adjacent frames are respectively calculated by difference value, and the moving amplitude in two directions is recorded. If the X axis change amplitude of a certain point is 2 pixels and the Y axis is 2 pixels, the position of the point in the plane is shifted. This process is performed for all feature points one by one to generate a complete difference matrix. Then the change trend of each feature point in the continuous frame is counted. Whether the change direction (such as the X coordinate of the continuous three frames is offset to the right) and the amplitude change are in the stable interval is judged. The stable interval is set to be within 3 pixels of the coordinate difference between each frame, and the change direction does not reverse continuously. If the proportion of the feature point that meets this condition within 10 frames is more than 80%, the trend of the point is considered stable. Further, the index of all trend-stable feature points is established. The trajectory curve of each point is drawn according to the planar coordinates changing with time. The change direction, fluctuation amplitude and continuity marking information of each curve are extracted. The abnormal change segment is marked and excluded. The remaining continuous change part is numbered and saved to the trend set. The trend information of all feature points is organized in frames to form the XY axis trend sequence reflecting the characteristics of the key points in the planar direction evolving with time.

[0099] S312: Based on the XY axis trend sequence, the distribution of the Z axis coordinates of each group of feature points in the continuous frame is detected, the continuous coordinates of the Z axis in the time sequence are combined, the mutation segment is recognized, the interpolation smoothing processing is performed, and the formula is adopted:

[0100] ;

[0101] The trajectory change parameters of each group of feature points in the three-axis space are obtained, and the abnormal trajectory is corrected to obtain the Z axis trajectory correction set, wherein, represents the trajectory change parameter of the group of feature points, represents the Z axis coordinate of the group of feature points in the frame, represents the Z axis coordinate of the group of feature points in the frame, represents the X-axis coordinate of the feature point in the first group of frames, represents the X-axis coordinate of the feature point in the first group of frames, represents the Y-axis coordinate of the feature point in the first group of frames, represents the Y-axis coordinate of the feature point in the first group of frames, represents the Y-axis coordinate of the feature point in the first group of frames, represents the Y-axis coordinate of the feature point in the first group of frames, represents the total number of frames in the time sequence;

[0102] The trajectory change parameter refers to a comprehensive parameter obtained by measuring the spatial trajectory fluctuation and change amplitude of the same group of feature points in three-dimensional space (X, Y, Z three-axis) in continuous frames over time, which is used to reflect the stability and variation degree of the spatial trajectory of the feature point in the entire time sequence. This parameter is used to identify and correct abnormal trajectory segments.

[0103] Detect the distribution of the Z-axis coordinate of each group of feature points in continuous frames, obtain the three-dimensional space coordinates of each group of feature points in a five-frame sequence, and construct the corresponding space trajectory sequence , wherein , derived from the positioning coordinates of two-dimensional image feature points, then according to the pixel depth mapping data obtained by the multi-spectral depth sensor, the corresponding extraction is carried out after normalization processing, and the subsequent operation is carried out in sequence. The Z-axis depth change and X-axis lateral displacement between adjacent frames are calculated, and the Y-axis interframe movement amplitude is used as the combination denominator item. Set the feature point number , the sampling results in a 5-frame time sequence are as follows:

[0104] The Z-axis original value is:

[0105] [138, 142, 146, 150, 149];

[0106] The X-axis is:

[0107] [62, 64, 66, 67, 66];

[0108] The Y-axis is:

[0109] [48, 47, 45, 44, 43];

[0110] The corresponding data after normalization processing is:

[0111] : [1.380, 1.420, 1.460, 1.500, 1.490];

[0112] : [0.620, 0.640, 0.660, 0.670, 0.660];

[0113] : [0.480, 0.470, 0.450, 0.440, 0.430];

[0114] Substitute the formula and calculate item by item as follows:

[0115] ;

[0116] ;

[0117] ;

[0118] ;

[0119] The feature point is calculated as The trajectory change parameter in a 5-frame time period is , if the upper limit of the preset trajectory correction range is 0.05, the parameter is in the correctable area, which indicates that the spatial change amplitude of the feature point in the observation time period is acceptable, there is a certain degree of jumping but not enough to constitute a trajectory break, interpolation correction can be performed, and then the local fluctuation area is smoothed to generate a continuous Z-axis trajectory correction set for subsequent three-dimensional trajectory sequence construction steps.

[0120] S313: Based on the Z-axis trajectory correction set, in combination with the trajectory continuity of the feature point in the three-axis space, a trajectory segment with balanced spatial change amplitude is screened, its continuous spatial coordinates in the time sequence are obtained, and a three-dimensional trajectory sequence is obtained;

[0121] Each feature point trajectory has been subjected to abnormal change point elimination and trajectory reconstruction operation, the time sequence trajectory data of the feature points in the three-dimensional coordinate system is read in turn, each trajectory line is analyzed in order of feature point number, the X, Y and Z direction coordinates recorded in each frame are traversed horizontally, the change amplitude difference in three directions between consecutive frames is recorded, and then it is judged whether the three-axis change amplitude of the point in the continuous time is stable, the judgment criterion is set as: if the coordinate change difference of each frame and the previous frame in each axis is not more than 4 pixels in 5 consecutive data, it is determined that the trajectory change amplitude is balanced, and is marked as a candidate trajectory segment, all candidate trajectory segments are numbered and summarized, if the length of a single segment is not less than 7 frames, and the three-axis coordinate change ratio does not deviate from the reference ratio ± 20%, the segment is included in the effective trajectory sequence, and further the timeline reconstruction is carried out on all effective segments, the X, Y and Z direction coordinates of each frame in the sequence are extracted, the point column array structure in three-dimensional space is constructed, and the three-dimensional trajectory sequence is saved as the basic data for subsequent behavior matching and action recognition.

[0122] Please refer to Figure 5 The action linkage matching data acquisition step is specifically:

[0123] S411: Based on the three-dimensional trajectory sequence, the X-axis and Y-axis position data of each frame feature point are selected, the continuous time sequence of the Z-axis trajectory is combined, the head rotation angle change amount is extracted, the angle difference of each frame face region displacement trajectory is calculated, the posture transformation angle under each time sequence is established with the corresponding feature point coordinate offset, and the angle displacement pairing data is obtained;

[0124] Each feature point has X-axis, Y-axis and Z-axis coordinate values ​​in continuous frames. For all feature points in each frame image, first extract the plane position of the X-axis and Y-axis. By setting the facial center point as the reference point, standardize the position offset direction and amplitude of each feature point. Then, combine the curve of the change of the Z-axis coordinate over time to judge the rotation relationship between the displacement path of each feature point in three-dimensional space and the overall posture of the head. Set the reference axis of the head posture change as the Z-axis vertical line. Take 3 frames as the time window, extract the feature point coordinates of the starting frame, middle frame and end frame respectively, and calculate the rotation angle change of the line connecting any two feature points in the three frames by the vector angle method. The angle change values ​​obtained from all feature points are averaged to form the overall posture transformation angle of the frame. It is judged whether it rises or falls continuously in the time series. If there are three consecutive frames with the same angle change trend, the trend change is recorded as a valid rotation sequence. The rotation angle change data is then paired with the spatial coordinate difference of the corresponding feature point. The posture change angle value in each frame is combined with the offset of the key feature point in the frame one-to-one. The corresponding frame number, angle value, corresponding point number and its three-dimensional offset value are recorded to form a pairing data set containing the angle displacement mapping relationship. The time axis data structure is constructed with the frame number as the main index to obtain the angle displacement pairing data.

[0125] S412: Based on the angle displacement pairing data, detect the micro-motion speed information of each area, compare the coordinate change amplitude and time interval of each area, and use the formula:

[0126] ;

[0127] Calculate rate joint offset trend ,in, Indicates the total number of regions involved in the comparison, Indicates the Micro-motion speed data obtained from area detection, Indicates the The coordinate change range of the regional feature points, Indicates the The time interval between adjacent frames in a region;

[0128] The rate joint deviation trend is the overall deviation between the micro-motion speed of facial feature points and the spatial motion speed in all comparison areas, which is used to reflect the consistency or coordination between multi-region motion and coordinate changes.

[0129] For the five facial regions defined in the surveillance scene, the micro-motion speed information recorded by the local motion capture unit in each region is detected one by one. The spatial coordinate change amplitude and time interval of each key feature point are calculated by combining the previous and next frame image data. The micro-motion speed is compared with the unit time coordinate displacement rate of the region, and the rate combined with the offset trend formula is used to calculate the offset of the above data, where: , indicating that there are five comparison areas. Indicates the The micro-movement speed of the area, in mm / s, Indicates the coordinate change range of the area between two consecutive frames, in mm. It represents the time interval between frames, in seconds. Now assume that the following raw data are obtained for regions 1 to 5 respectively:

[0130] 、 、 、 、 ;

[0131] The corresponding displacement amplitude and time interval are:

[0132] 、 、 、 、 (All units are in mm);

[0133] 、 、 、 、 (All units are in seconds);

[0134] After normalizing the units of each region, the following rate terms are obtained:

[0135] The first area is , ;

[0136] The second area is , ;

[0137] The third area is , ;

[0138] The fourth area is , ;

[0139] The 5th area is , ;

[0140] After substituting the formula respectively, we can get:

[0141] ;

[0142] ;

[0143] The results show that the rate joint offset trend calculated by the current calculation , compared with the set action linkage offset judgment reference interval [0, 1.2], is within the allowable matching tolerance range, which means that the overall difference between the micro-motion speed and the feature point displacement rate in the five regions does not exceed the upper limit boundary of the judgment standard, and the linkage response consistency is strong, which means that all regions currently have strong coordination in structural motion synchronization, which can be used as a reference basis for the next action and structure matching judgment, and used to assist in establishing the effective combination sequence in the action linkage matching data.

[0144] S413: Based on the rate joint offset trend, data comparison is performed on each group of micro-motion data and feature point spatial motion change, the correlation degree of the structure similarity parameters and the action linkage index in the combination is analyzed, the combination that meets the linkage condition is collected, and the action linkage matching data is obtained;

[0145] The corresponding time stamps of each group of action data and feature point spatial change data captured by the facial micro-motion recording unit are obtained, the micro-motion time sequence and three-dimensional trajectory data are aligned through synchronous processing, the associated key action nodes in each group of micro-motion events are extracted first, such as mouth up, eyebrow down, etc., and the spatial coordinate offset value and motion direction of the feature points in the three frames before and after the action occurs are extracted, the offset rate in X, Y and Z directions is recorded respectively, the spatial feature point number corresponding to each group of micro-motion data is bound, whether it is consistent with the trend direction in the feature point trajectory under the same time sequence is counted, if the feature point coordinate offset direction corresponding to the micro-motion occurrence frame is consistent with the trend before and after the frame, and the change rate is within the set threshold ± 15%, then record this group as a linkage established sample, then the structure similarity of the feature point distribution form in all samples is analyzed, the difference between the relative distance relationship of the feature points and the standard face structure model is counted, if the average value of the difference is less than the set structure similarity threshold 3 pixels, and the area ratio of the micro-motion occurrence region and the linkage feature point distribution overlap area is more than 80%, then the data combination is identified as meeting the linkage condition, finally all the sample combinations that meet the above offset rate matching, direction consistency and structure similarity are extracted, four types of information items of action number, time stamp, linkage point number set and spatial offset value are generated, and are sorted into action linkage matching data.

[0146] Please refer to Figure 6 , the acquisition steps of the face recognition mapping result are as follows:

[0147] S511: Based on the action linkage matching data, the correspondence between the numbered feature points and the behavior parameters is analyzed, the continuous distribution of each feature point in the three-dimensional space and the synchronicity of the behavior change are judged, the corresponding groups with high consistency of structural features and behavior features are screened, and the corresponding distribution data is obtained;

[0148] The data contains the number of each group of feature points and the corresponding behavior parameter sequence. First, the continuous frame coordinate trajectories of all feature points in the three-dimensional space are extracted, a time sequence index table is established for the X, Y, Z three-axis coordinates of each point, the trajectory paragraphs of each group of feature points within 10 consecutive frames are extracted, and the inter-frame position change vector is recorded. Then, the behavior parameter data at the same time sequence is positioned synchronously, the change markers of action parameters such as "eyebrow lifting", "mouth corner offset", "head left turn" are extracted, and whether the trajectory change direction of each feature point is synchronized with the time sequence change of the behavior parameter is analyzed. For example, if the Y-axis coordinate of feature point number 23 continuously rises and the amplitude exceeds 3 pixels within the 5th frame to the 7th frame, and the action marked during this period is eyebrow lifting, it is determined that the number 23 and the behavior parameter have synchronicity. The above judgment is performed on all numbered feature points, the numbers that pass the synchronization judgment are marked, and the pairing relationship between them and the action label is recorded. Further, the synchronous pairs are screened for structural consistency. By counting whether the change difference of the three-axis motion trajectory in all effective frames is controlled within the threshold range, if the motion direction of the continuous three frames remains consistent and the offset value difference of each frame does not exceed 4 pixels, it is determined that the structural change of this group is stable. The numbered behavior pair group with this feature is retained in the screening result set. Finally, a mapping table of one-to-one correspondence between numbered feature points and behavior parameters is constructed, and the corresponding group with high consistency of structure and behavior parameters is obtained, forming the corresponding distribution data.

[0149] S512: Based on the corresponding distribution data, the spatial structural relationship and the behavior change mode of the difference data group are compared, the arrangement order of the feature point combination and the corresponding rules between the behavior parameters are optimized, the combination with consistent structure and behavior is screened, and the structural mapping combination data is obtained;

[0150] The selected feature points and their corresponding behavior parameters are further divided into different sub-combinations. By comparing the relative coordinate structure of the feature points in each group in three-dimensional space, it is determined whether there is a shape deviation between the point groups. The relative distance of the feature points with the same number in the two groups in the three-axis space is calculated, and then it is compared whether the difference exceeds the set structure similarity threshold of 5 pixels. For example, the average value of feature point group A in the Z-axis direction is 24, and the average value of group B is 29. The structure deviation is 5. If the difference is greater than the threshold, it is marked as a difference combination. Then, the change mode of the behavior parameters corresponding to the difference combination is analyzed to determine whether the trend of the parameters in the frame sequence is consistent with the original combination. For example, the behavior parameter in group A is continuously rising, while the behavior parameter in group B is intermittently changing. It is recorded as inconsistent behavior mode. Finally, all structure difference and behavior difference combinations are classified. The combinations with a structure deviation less than 5 pixels and consistent behavior change trend are marked as preferred combinations. The arrangement order of the feature points in the preferred combinations is reconstructed, and the three-axis coordinate stability priority is sorted. The number order is adjusted and the new arrangement rule is recorded. The behavior parameters are re-grouped according to their correlation with the change trend of the rearranged feature points. The pairing relationship with a correlation greater than 0.85 is retained, and the rest is excluded. Finally, the structure and behavior parameter combination pairs with consistent structure and behavior are output, and the structure mapping combination data is generated.

[0151] S513: Based on the structure mapping combination data, the linkage distribution between the feature point groups and the behavior parameter groups is determined. The participation ratio of each group of feature points and the linkage characteristics of the behavior parameters are calculated. The synchronization relationship and mapping index are marked to obtain the face recognition mapping result.

[0152] The number of each group of feature points and the corresponding behavior parameters is linked to determine the relationship. First, the ratio of the number of feature points in each group to the total number of feature points is extracted. Let the ratio be the participation ratio. For example, if the total number of feature points is 40 and the current combination involves 28 numbers, the participation ratio is 70%. The change frequency and amplitude of each group of behavior parameter sequences in the action trigger frame segment are extracted. It is determined whether they have a spatial deviation in the same frame. If more than 80% of the points move in the same direction in the same frame, and the frame is marked as a specific action change frame, it is marked as a strong linkage relationship. Then, all feature point combinations are counted. The occurrence frequency, behavior consistency rate, and participation ratio of each group are calculated. Set the participation ratio to be higher than 60% and the behavior consistency rate to be higher than 85% as the screening standard. All combinations that meet the conditions are marked as synchronous linkage relationship established groups. Then, the mapping index number is generated according to the feature point number set and the behavior parameter sequence. For example, feature point group T17-32 is linked to action groups A2 and A5. The index number is marked as IDX_T17-32_A2A5. The index number is used as the final recognition label output. The number correspondence table and the action parameter association table are formed, and the synchronization label state is marked. The face recognition mapping result with a unique identification is generated.

[0153] The above merely describes the preferred embodiments of the present application, but is not intended to limit the present application in other forms. Any person skilled in the art can make changes or modifications to the above disclosed technical contents into equivalent embodiments with equivalent changes, and apply to other fields. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application without departing from the technical solution content of the present application still belongs to the protection scope of the present application technical solution.

Claims

1. An AI-based face recognition security method, characterized in that: The following steps are involved: S1: Based on the monitoring entrance and exit recognition channel, analyze the pixel distribution of the facial image, determine the contour clarity captured by the infrared vision array, compare the acquisition details of the multispectral sensor, screen the feature points and symmetry parameters of the complete frame, determine the skin color area distribution, and obtain the multispectral facial features; S2: Based on the multispectral facial features, calling the image recognition buffer unit, optimizing the multi-frame images of the facial feature aggregation channel, judging the uniformity of the spatial distribution of feature points, comparing the overlap of contour structures, screening consecutive facial frames with spatial differences within an interval, and obtaining a spatial feature screening sequence; S3: Based on the spatial feature screening sequence, analyze the three-dimensional spatial changes of feature points, compare the X-axis and Y-axis position trends of consecutive frames, determine the Z-axis trajectory continuity, screen feature points with stable position changes, perform corrections on trajectory mutation points, and obtain a three-dimensional trajectory sequence; S4: Based on the three-dimensional trajectory sequence, calculating the change in head rotation angle, determining the speed of facial micro-movements recorded by the local motion capture unit, comparing the movement parameters with the spatial coordinate change pattern of the feature points, selecting the data with the best matching degree, and obtaining the movement linkage matching data; S5: Based on the action linkage matching data, the correspondence between the numbered feature points and the behavior parameters is analyzed, the consistency of the data structure of each group of recognition channels is determined, the joint recognition integrated port discrimination process is optimized, the group with both structure and behavior meeting the standards is identified, the identification number is marked, and the face recognition mapping result is obtained; The face recognition mapping result includes an identity mapping identifier, an identification identifier, and a feature linkage mapping index.

2. The AI-based face recognition security method according to claim 1, characterized in that: The multispectral facial features include spectral distribution parameters, texture features, and skin color block information; the spatial feature screening sequence includes spatial consistency indicators, structural aggregation labels, and inter-frame stable segments; the three-dimensional trajectory sequence includes coordinate trajectory data, coherence marks, and corrected trajectory sets; and the action linkage matching data includes behavior linkage parameters, action identification numbers, and linkage matching relationship groups.

3. The AI-based face recognition security method according to claim 1, characterized in that: The steps of acquiring the multispectral facial features are specifically as follows: S111: Based on the monitoring entrance and exit recognition channel, the pixel distribution of the facial image data is analyzed. For each block of the image sequence, the distribution density of the edge continuity of each pixel block is calculated by statistically analyzing the change level of the grayscale gradient. The edge clarity change trend of each area in the infrared vision array capture area is determined to obtain the infrared contour clarity distribution; S112: Based on the infrared contour clarity distribution, compare the brightness histogram distribution in the images acquired by the multi-spectral synchronous sensor array at different time points, calculate the density of texture directions in each area, divide the facial key areas by analyzing the aggregation and closure characteristics of boundary pixels, and obtain facial structure aggregation data; S113: Based on the facial structure aggregated data, feature point coordinates and symmetry mapping sets in the image frame are screened, and multi-channel fusion and superposition are performed to obtain multispectral facial features.

4. The AI-based face recognition security method according to claim 1, characterized in that: The steps for obtaining the spatial feature screening sequence are specifically as follows: S211: Based on the multispectral facial features, analyzing the arrangement of each feature point on the horizontal and vertical axes, determining whether its spatial distribution has the balanced and symmetrical characteristics of the facial structure, selecting image frames with consistent spatial distribution patterns, and obtaining a set of symmetrically distributed image frames; S212: Based on the symmetrically distributed image frame set, comparing the spatial orientation and connection relationship of the boundary lines in the same area, calculating the continuity of key contour lines and boundary connection between frames, and selecting frame images with consistent contour connection patterns to obtain a contour-continuous image frame set; S213: Based on the contour continuous image frame set, determine the coordinate changes of the feature points of each frame, analyze the movement trend of the key points between adjacent image frames, and select image frame sequences with stable feature point movement patterns to obtain a spatial feature screening sequence.

5. The AI-based face recognition security method according to claim 1, characterized in that: The steps for obtaining the three-dimensional trajectory sequence are specifically as follows: S311: Based on the spatial feature screening sequence, extract the plane coordinates of each feature point in the X-axis and Y-axis in consecutive frames, compare the coordinate change amplitude of the same feature point in adjacent frames, identify its change trend in the time series, and establish an XY axis trend sequence; S312: Based on the XY axis trend sequence, the distribution of the Z axis coordinates of each set of feature points in the continuous frames is detected. Combined with the continuous Z axis coordinates in the time series, the mutation segments are identified, and interpolation and smoothing processing is performed to obtain the trajectory change parameters of each set of feature points in the three-axis space, and the abnormal trajectory is corrected to obtain a Z axis trajectory correction set; S313: Based on the Z-axis trajectory correction set and in combination with the trajectory continuity of the feature points in the three-axis space, the trajectory segments with balanced spatial variation amplitudes are screened, and their continuous spatial coordinates in the time series are obtained to obtain a three-dimensional trajectory sequence.

6. The AI-based face recognition security method according to claim 1, characterized in that: The steps for obtaining the action linkage matching data are specifically as follows: S411: Based on the three-dimensional trajectory sequence, the X-axis and Y-axis position data of the feature points of each frame are selected, and combined with the continuous time series of the Z-axis trajectory, the head rotation angle change data is extracted, and the angle difference of the facial region displacement trajectory of each frame is calculated. The posture transformation angle in each time series is paired with the corresponding feature point coordinate offset to obtain angle displacement paired data; S412: Based on the angle displacement paired data, detecting micro-motion speed information of each region, comparing the coordinate change amplitude and time interval of each region, and calculating the rate joint offset trend; S413: Based on the rate joint offset trend, perform data comparison on each group of micro-motion data and the spatial motion changes of the feature points, analyze the correlation between the structural similarity parameters and the motion linkage indicators in the combination, collect data on the combinations that meet the linkage conditions, and obtain motion linkage matching data.

7. The AI-based face recognition security method according to claim 1, characterized in that: The steps for obtaining the face recognition mapping result are specifically as follows: S511: Based on the action linkage matching data, analyzing the correspondence between the numbered feature points and the behavior parameters, determining the synchronization between the continuous distribution of each feature point in the three-dimensional space and the behavior change, screening corresponding groups with highly consistent structural features and behavioral features, and obtaining corresponding distribution data; S512: Based on the corresponding distribution data, compare the spatial structure relationship and behavior change pattern of the difference data group, optimize the correspondence rules between the arrangement order of the feature point combination and the behavior parameters, select the combination with consistent structure and behavior, and obtain structure mapping combination data; S513: Based on the structure mapping combination data, determine the linkage distribution between the feature point group and the behavior parameter group, calculate the participation ratio of each group of feature points and the linkage characteristics of the behavior parameters, mark the synchronization relationship and mapping index, and obtain the face recognition mapping result.

Images